I can't edit any more, so I'm posting as a reply. I compared my dataset and their data, and it turns out that even in Brazil there is significant correlation b

I can't edit any more, so I'm posting as a reply. I compared my dataset and their data, and it turns out that even in Brazil there is significant correlation between mother education and breastfeeding, barely less so than on the dataset I have. So you should probably disregard my edit in the above post: accounting for confounders could be important in both cases.